Re: st: goodness of fit svy: probit

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Thu, 10/2/11, Bart van Rijsbergen wrote me privately:
> Thanks for your response.
> See the do- and logfile.
> I'm using stata/SE 10.1

On Statalist you respond to the list not to individuals. 
<http://www.stata.com/support/faqs/res/statalist.html#private>
Also note that on the list attachments are not allowed.
<http://www.stata.com/support/faqs/res/statalist.html#toask>

-help probit postestimation- under Stata 10 says that 
-estat gof- is not allowed when the -svy- prefix was
used. So, you can either update to Stata 11, or not 
use these tests. I don't think these tests are good 
idea anyhow, so my advise would be just to forget 
about it.

When we are doing model selection we are making a 
trade-off between simplicity/usefulness on the one hand 
and fit/realism on the other. A test is not a useful 
operationalisation of that trade-off. However, we are 
so used to using tests all the time that in many 
disciplines we think that we should also use it for 
model selection to make it more "scientific". Basically, 
we are suffering from the WYHAHAPLLN-syndrom (when-you-
have-a-hammer-all-problems-look-like-nails-syndrom (I 
think I still need to work on a more catchy acronym)).

-- Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/